Automated Smart Attendance System for Higher Education using a Lightweight DNN Model-based Face Recognition

This research introduces a novel online system for automated student attendance utilising modern technologies such as Streamlit, Redis, Python, and facial recognition. Conventional attendance methods, like human roll calls and barcode scanning, are susceptible to inaccuracies, deceit, and inefficiency. The suggested system utilises deep learning methodologies to develop a resilient facial recognition model, offering a precise and efficient alternative. Feature extraction is performed via the Local Binary Patterns Histogram (LBPH) technique, and the model is executed with the DS-CNN-LSTM architecture to improve efficacy. Validation on the Smart Attendance dataset confirms the system's efficacy, attaining a maximum training recognition rate of 97.56% and a validation accuracy of 95.32%. Cross-validation and temporal complexity analyses substantiate the model's dependability and computational efficacy. This method provides a reliable, efficient, and fraud-resistant alternative for attendance management in educational environments.